T-Rex 0.26.8 ETHW, ETC (+dual mining) GPU miner - page 105.

trexminer

member

Activity: 283

Merit: 63

Quote from: nitrobg on June 10, 2018, 03:26:28 AM

Maybe it is a too high overclock, I really don't know as my rig is mixed with many different cards. tpruvot's ccminer worked fine, though. I've lowered to overclock on all cards that ran past 2050MHz, we'll see if crashes are still there.
The default ccminer api would be lovely, so the hashrate monitoring would work on Hive.

Lyra2z doesn't show improvement over high intensities, but it works up to -i 26 on 1080 Ti.
Phi however gets a 4% boost from a higher intensity. Testing on 1080 Ti @ 200W, -1000 memory, +120 core.
-i 20 (default): 37.8-38.1MH/s
-i 27: 39.5MH/s

As I don't own a 1080 Ti, this is very valuable info, once again thanks for that. I'll see what I can do about the API.
I know it might be a lot to ask but could you also benchmark Tribus algo with your 1080 Ti that I just added?

trexminer

member

Activity: 283

Merit: 63

Added Tribus algo. See download links in the first post.

nitrobg

member

Activity: 413

Merit: 17

Quote from: trexminer on June 09, 2018, 10:19:18 PM

Many thanks for reporting this. Unfortunately I couldn't replicate the crash you've experienced, tried it on 1050, 1070 and 1070ti and it all worked fine with -i 17. Was it 1060 or 1080ti that crashed?
The API, along with the windows version, should be available in a couple of weeks.
Also, I was wondering if you could benchmark phi1612 on your 1080ti for our stats?

Maybe it is a too high overclock, I really don't know as my rig is mixed with many different cards. tpruvot's ccminer worked fine, though. I've lowered to overclock on all cards that ran past 2050MHz, we'll see if crashes are still there.
The default ccminer api would be lovely, so the hashrate monitoring would work on Hive.

Lyra2z doesn't show improvement over high intensities, but it works up to -i 26 on 1080 Ti.
Phi however gets a 4% boost from a higher intensity. Testing on 1080 Ti @ 200W, -1000 memory, +120 core.
-i 20 (default): 37.8-38.1MH/s
-i 27: 39.5MH/s

trexminer

member

Activity: 283

Merit: 63

Quote from: nitrobg on June 09, 2018, 09:30:42 AM

3750MH/s on 1080Ti
1200MH/s on 1060/3

Hive doesn't have the libcudart.so.9.1, so perhaps you should include the library in the binary.

The miner crashes on the default intensity of 17 with this error:

Code:

*** Error in `./t-rex': free(): invalid size: 0x00007fc12c68e1d0 ***
======= Backtrace: =========
/lib/x86_64-linux-gnu/libc.so.6(+0x777e5)[0x7fc1834647e5]
/lib/x86_64-linux-gnu/libc.so.6(+0x8037a)[0x7fc18346d37a]
/lib/x86_64-linux-gnu/libc.so.6(cfree+0x4c)[0x7fc18347153c]
/usr/lib/x86_64-linux-gnu/libcuda.so.1(+0x1ebd3c)[0x7fc17df70d3c]
/usr/lib/x86_64-linux-gnu/libcuda.so.1(+0xf886b)[0x7fc17de7d86b]
/usr/lib/x86_64-linux-gnu/libcuda.so.1(+0x248808)[0x7fc17dfcd808]
/usr/lib/x86_64-linux-gnu/libcuda.so.1(+0x1e4e80)[0x7fc17df69e80]
/lib/x86_64-linux-gnu/libpthread.so.0(+0x76ba)[0x7fc1849186ba]
/lib/x86_64-linux-gnu/libc.so.6(clone+0x6d)[0x7fc1834f441d]
======= Memory map: ========
00400000-00595000 r-xp 00000000 00:00 0
00595000-00794000 ---p 00000000 00:00 0
00794000-00795000 r-xp 00000000 00:00 0
00795000-007bf000 rwxp 00000000 00:00 0
00800000-00801000 rwxp 00000000 00:00 0
00fd0000-01064000 rwxp 00000000 00:00 0                                  [heap]
200000000-200200000 rwxs 00000000 00:06 15577                            /dev/nvidiactl
200200000-200400000 rwxs 00000000 00:06 15577                            /dev/nvidiactl
200400000-200600000 rwxs 00000000 00:06 15577                            /dev/nvidiactl
200600000-200800000 rwxs 00000000 00:06 15577                            /dev/nvidiactl
200800000-200a00000 rwxs 00000000 00:06 15577                            /dev/nvidiactl
200a00000-200c00000 rwxs 00000000 00:06 15577                            /dev/nvidiactl
...........

Lowering it down to intensity 16 seems to have solved it.
Can you enable the default API?

Many thanks for reporting this. Unfortunately I couldn't replicate the crash you've experienced, tried it on 1050, 1070 and 1070ti and it all worked fine with -i 17. Was it 1060 or 1080ti that crashed?
The API, along with the windows version, should be available in a couple of weeks.
Also, I was wondering if you could benchmark phi1612 on your 1080ti for our stats?

trexminer

member

Activity: 283

Merit: 63

Quote from: pileto3 on June 09, 2018, 09:10:08 AM

Any chances for Windows version?

We estimate a windows version to be available in a couple of weeks. There's two of us working on this miner, and windows version + API are being developed in parallel to adding new algos.

nitrobg

member

Activity: 413

Merit: 17

3750MH/s on 1080Ti
1200MH/s on 1060/3

Hive doesn't have the libcudart.so.9.1, so perhaps you should include the library in the binary.

The miner crashes on the default intensity of 17 with this error:

Code:

*** Error in `./t-rex': free(): invalid size: 0x00007fc12c68e1d0 ***
======= Backtrace: =========
/lib/x86_64-linux-gnu/libc.so.6(+0x777e5)[0x7fc1834647e5]
/lib/x86_64-linux-gnu/libc.so.6(+0x8037a)[0x7fc18346d37a]
/lib/x86_64-linux-gnu/libc.so.6(cfree+0x4c)[0x7fc18347153c]
/usr/lib/x86_64-linux-gnu/libcuda.so.1(+0x1ebd3c)[0x7fc17df70d3c]
/usr/lib/x86_64-linux-gnu/libcuda.so.1(+0xf886b)[0x7fc17de7d86b]
/usr/lib/x86_64-linux-gnu/libcuda.so.1(+0x248808)[0x7fc17dfcd808]
/usr/lib/x86_64-linux-gnu/libcuda.so.1(+0x1e4e80)[0x7fc17df69e80]
/lib/x86_64-linux-gnu/libpthread.so.0(+0x76ba)[0x7fc1849186ba]
/lib/x86_64-linux-gnu/libc.so.6(clone+0x6d)[0x7fc1834f441d]
======= Memory map: ========
00400000-00595000 r-xp 00000000 00:00 0
00595000-00794000 ---p 00000000 00:00 0
00794000-00795000 r-xp 00000000 00:00 0
00795000-007bf000 rwxp 00000000 00:00 0
00800000-00801000 rwxp 00000000 00:00 0
00fd0000-01064000 rwxp 00000000 00:00 0                                  [heap]
200000000-200200000 rwxs 00000000 00:06 15577                            /dev/nvidiactl
200200000-200400000 rwxs 00000000 00:06 15577                            /dev/nvidiactl
200400000-200600000 rwxs 00000000 00:06 15577                            /dev/nvidiactl
200600000-200800000 rwxs 00000000 00:06 15577                            /dev/nvidiactl
200800000-200a00000 rwxs 00000000 00:06 15577                            /dev/nvidiactl
200a00000-200c00000 rwxs 00000000 00:06 15577                            /dev/nvidiactl
...........

Lowering it down to intensity 16 seems to have solved it.
Can you enable the default API?

pileto3

newbie

Activity: 14

Merit: 0

Any chances for Windows version?

trexminer

member

Activity: 283

Merit: 63

Quote from: Berni5813 on June 09, 2018, 08:00:24 AM

Is there any chance to get it compiled for cuda 9.2 or cuda 8?

Thank you! nice work

Here you go, cuda 9.2 version - https://mega.nz/#!qJsQlAKB!rOpHtsqWjwvKzIljvL1uXO4bv3G-ZeWEB1XAMLWvpgk
I haven't tested it properly though, may not be stable or as fast as 9.1. Let me know how it goes.

Berni5813

newbie

Activity: 3

Merit: 0

Is there any chance to get it compiled for cuda 9.2 or cuda 8?

Thank you! nice work

trexminer

member

Activity: 283

Merit: 63

Just added Lyra2z support. Download link is at the top.

trexminer

member

Activity: 283

Merit: 63

Quote from: Kompik on June 08, 2018, 02:12:17 PM

Lux should be changing algo soon, is this miner compatible with the new algo?

LUX is going to switch to phi2 algo which uses different crypto functions under the hood. I'm planning on adding phi2 support but not until I make it fast enough, otherwise there's no point.

Kompik

sr. member

Activity: 463

Merit: 250

Lux should be changing algo soon, is this miner compatible with the new algo?

trexminer

member

Activity: 283

Merit: 63

In fact, I just found a person who faced exactly the same issue and the solution was "sudo apt install libcurl3" (Ubuntu 16.04)

trexminer

member

Activity: 283

Merit: 63

Quote from: minimumnz on June 07, 2018, 08:59:46 PM

Getting this error on a couple machines one Ubuntu 16.04 the other 18.04

./t-rex: /usr/lib/x86_64-linux-gnu/libcurl.so.4: version `CURL_OPENSSL_3' not found (required by ./t-rex)

I'm guessing libcurl3 package isn't up to date, at least on 16.04. Could you contact me via direct messages please? We'll try to investigate this issue. I can't send you a direct message because of the default bitcointalk policy not to accept them from newbies.

minimumnz

newbie

Activity: 2

Merit: 0

Getting this error on a couple machines one Ubuntu 16.04 the other 18.04

./t-rex: /usr/lib/x86_64-linux-gnu/libcurl.so.4: version `CURL_OPENSSL_3' not found (required by ./t-rex)

trexminer

member

Activity: 283

Merit: 63

Quote from: Freew0rld on June 07, 2018, 01:36:43 PM

Windows support?

Not yet, but it's certainly on my to-do list.

trexminer

member

Activity: 283

Merit: 63

Quote from: nitrobg on June 07, 2018, 10:52:35 AM

1060/3@75W: 12.8MH/s
1060/6@75W: 14MH/s

Thank you, around 5% hashrate improvement over Enemy and anxmod. Hope it is the same on poolside.

Thanks for reporting that. I'm sure others will find your results very helpful.

Freew0rld

newbie

Activity: 9

Merit: 0

Windows support?

nitrobg

member

Activity: 413

Merit: 17

1060/3@75W: 12.8MH/s
1060/6@75W: 14MH/s

Thank you, around 5% hashrate improvement over Enemy and anxmod. Hope it is the same on poolside.

trexminer

member

Activity: 283

Merit: 63

Unfortunately I don't have any 1080TIs. If you do, it would be really great if you could try it and post the results. Thanks!

Topic: T-Rex 0.26.8 ETHW, ETC (+dual mining) GPU miner - page 105. (Read 175032 times)